
%run "/code/source/config/notebook_settings.py"


from source.library.text_analysis import count_tokens, tf_idf, get_context_from_keyword, count_keywords, \
    count_keywords_by, impurity


with Timer("Loading Data"):
    path = 'artifacts/data/processed/un-general-debates-blueprint.pkl'
    df = pd.read_pickle(path)

2023-02-26 22:48:04 - INFO     | Timer Started: Loading Data
2023-02-26 22:48:06 - INFO     | Timer Finished: (2.29 seconds)


hlp.pandas.numeric_summary(df)


hlp.pandas.non_numeric_summary(df)


df.head(2)


df['text'].apply(impurity).max()

0.0023444160272804776


df['text'].iloc[0][0:1000]

'33: May I first convey to our President the congratulations of the Albanian delegation on his election to the Presidency of the twenty-fifth session of the General Assembly?\n34.\tIn taking up the work on the agenda of the twenty- fifth session of the General Assembly, which is being held on the eve of the twenty-fifth anniversary of the coming into force of the Charter of the United Nations, the peace-loving Member States would have wished to be in a position to present on this occasion some picture of positive and satisfactory activity on the part of the United Nations. The Albanian delegation, for its part, would have taken great pleasure in drawing up such a balance sheet of activities covering a quarter of a century, which is certainly no short period in the life of an international organization. Unfortunately, this is not the situation. Created on the day after victory had been achieved over the Powers of the Rome BerlinTokyo Axis and conceived in the spirit of the principles which'


'|'.join(df['tokens'].iloc[0])[0:1000]

'may|first|convey|president|congratulations|albanian|delegation|election|presidency|twenty-fifth|session|general|assembly|taking|work|agenda|twenty-|fifth|session|general|assembly|held|eve|twenty-fifth|anniversary|coming|force|charter|united|nations|peace-loving|member|states|would|wished|position|present|occasion|picture|positive|satisfactory|activity|part|united|nations|albanian|delegation|part|would|taken|great|pleasure|drawing|balance|sheet|activities|covering|quarter|century|certainly|short|period|life|international|organization|unfortunately|situation|created|day|victory|achieved|powers|rome|berlintokyo|axis|conceived|spirit|principles|predominated|war|antifascist|coalition|organization|awakened|whole|progressive|humanity|hope|would|serve|important|factor|creating|better|international|conditions|order|favor|cause|freedom|peace|world|security|activities|number|events|occurred|world|arena|period|disappointed|hopes|peoples|united|nations|far|contributed|required|fundamental|provision'


'|'.join(df['bi_grams'].iloc[0])[0:1000]

'first convey|albanian delegation|twenty-fifth session|general assembly|twenty- fifth|fifth session|general assembly|twenty-fifth anniversary|united nations|peace-loving member|member states|states would|satisfactory activity|united nations|albanian delegation|part would|taken great|great pleasure|balance sheet|activities covering|short period|international organization|organization unfortunately|situation created|rome berlintokyo|berlintokyo axis|antifascist coalition|organization awakened|progressive humanity|would serve|important factor|creating better|better international|international conditions|freedom peace|world security|world arena|period disappointed|united nations|nations far|fundamental provisions|international peace|liberation struggle|imperialist powers|united states|america foremost|foremost among|path diametrically|diametrically opposed|instrument favoring|pillage oppression|peace-loving peoples|united nations|committing aggression|many parts|frequently helped|direction '


df[df['speaker'].str.contains('Bush')]['speaker'].value_counts()

George W. Bush        4
Mr. George W. Bush    2
Bush                  1
George Bush           1
Mr. George W Bush     1
Name: speaker, dtype: int64


count_tokens(df['tokens']).head(20)


ax = df['text_length'].plot(kind='box', vert=False, figsize=(10, 1))
ax.set_title("Distribution of Text Length")
ax.set_xlabel("# of Characters")
ax.set_yticklabels([])
ax;


ax = df['text_length'].plot(kind='hist', bins=60, figsize=(10, 2));
ax.set_title("Distribution of Text Length")
ax.set_xlabel("# of Characters")
ax;


import seaborn as sns
sns.displot(df['text_length'], bins=60, kde=True, height=3, aspect=3);


where = df['country'].isin(['USA', 'FRA', 'GBR', 'CHN', 'RUS'])
g = sns.catplot(data=df[where], x="country", y="text_length", kind='box')
g.fig.set_size_inches(6, 3)
g.fig.set_dpi(100)
g = sns.catplot(data=df[where], x="country", y="text_length", kind='violin')
g.fig.set_size_inches(6, 3)
g.fig.set_dpi(100)


assert not df[['year', 'country']].duplicated().any()
df.groupby('year').size().plot(title="Number of Countries");


df.\
    groupby('year').\
    agg({'text_length': 'mean'}).\
    plot(title="Avg. Speech Length", ylim=(0,30000));


counts_df = count_tokens(df['tokens'])


counts_df.head()


def plot_wordcloud(frequency_dict):
    wc = wordcloud.WordCloud(
        background_color='white',
        #colormap='RdYlGn',
        colormap='tab20b',
        width=round(hlpp.STANDARD_WIDTH*100),
        height=round(hlpp.STANDARD_HEIGHT*100),
        max_words = 200, max_font_size=150,
        random_state=42
    )
    wc.generate_from_frequencies(frequency_dict)

    fig, ax = plt.subplots(figsize=(hlpp.STANDARD_WIDTH, hlpp.STANDARD_HEIGHT))
    ax.imshow(wc, interpolation='bilinear')
    #plt.title("XXX")
    plt.axis('off')


plot_wordcloud(counts_df.to_dict()['frequency']);


tf_idf_df = tf_idf(
    df=df,
    tokens_column='tokens',
    segment_columns = None,
    min_frequency_corpus=20,
    min_frequency_document=20,
)


ax = tf_idf_df.\
    sort_values(by='frequency', ascending=False).\
    head(30)[['frequency']].\
    plot(kind='barh', width=0.99)
ax.set_title("Frequency of Uni-Grams")
ax.set_xlabel("Frequency")
ax.invert_yaxis();


ax = tf_idf_df.head(30)[['tf-idf']].plot(kind='barh', width=0.99)
ax.set_title("TF-IDF of Uni-Grams")
ax.set_xlabel("TF-IDF")
ax.invert_yaxis();


plot_wordcloud(tf_idf_df.to_dict()['tf-idf']);


tf_idf_per_year = tf_idf(
    df=df,
    tokens_column='tokens',
    segment_columns = 'year',
    min_frequency_corpus=10,
    min_frequency_document=10,
)


stop_words = ['twenty-fifth', 'twenty-five', 'twenty', 'fifth']
tokens_to_show = tf_idf_per_year.query('year == 1970').reset_index()
tokens_to_show = tokens_to_show[~tokens_to_show.token.isin(stop_words)]


ax = tokens_to_show.head(30).set_index('token')[['tf-idf']].plot(kind='barh', width=0.99)
ax.set_title("TF-IDF of Uni-Grams - 1970")
ax.set_xlabel("TF-IDF")
ax.invert_yaxis();


tokens_to_show = tokens_to_show[['token', 'tf-idf']].set_index('token')
tokens_to_show = tokens_to_show.to_dict()['tf-idf']


plot_wordcloud(tokens_to_show);


stop_words = ['seventieth']
tokens_to_show = tf_idf_per_year.query('year == 2015').reset_index()
tokens_to_show = tokens_to_show[~tokens_to_show.token.isin(stop_words)]


ax = tokens_to_show.head(30).set_index('token')[['tf-idf']].plot(kind='barh', width=0.99)
ax.set_title("TF-IDF of Uni-Grams - 2015")
ax.set_xlabel("TF-IDF")
ax.invert_yaxis();


tokens_to_show = tokens_to_show[['token', 'tf-idf']].set_index('token')
tokens_to_show = tokens_to_show.to_dict()['tf-idf']


plot_wordcloud(tokens_to_show);


contexts = get_context_from_keyword(
    documents=df[df['year'] == 2015]['text'],
    window_width=50,
    keyword='sdgs', random_seed=42
)
for x in contexts:
    print(x)

-29946  03/10/2015 A/70/PV.28 Development Goals ( |SDGs| ). Beyond the immediate present, the Commonwealth 
nd collectively, to ensure the realization of the  |SDGs| . The Commonwealth of Dominica is therefore encour
that and other initiatives aimed at achieving the  |SDGs| . SIDS are equally concerned about sustainably uti
n of the post-2015 Sustainable Development Goals ( |SDGs| ). I should like to take this opportunity to conve
nian State. We must ensure the realization of the  |SDGs|  in a timely manner. We hope that the upcoming cli
s to translate the Sustainable Development Goals ( |SDGs| ) 15-29664 21/25  A/70/PV.23 01/10/2015 of the ne
w Agenda into action. That can happen only if the  |SDGs|  are matched with the necessary resources, in acco
, smart and innovatively to achieve the 17 global  |SDGs|  and the 169 targets. It has to be driven by a new
t for our rural population in accordance with the  |SDGs| . The leaders of the Pacific small island developi
of the gender-related Goals and targets under the  |SDGs| . In 2009, Solomon Islands, as an archipelagic Sta


contexts = get_context_from_keyword(
    documents=df[df['year'] == 2015]['text'],
    window_width=50,
    keyword='sids', random_seed=42
)
for x in contexts:
    print(x)

ntry, our region, small island developing States ( |SIDS| ) and our planet. We decided then, as we have done
 to our planet, with a disproportionate impact on  |SIDS| . Our concerns have been magnified since 27 August
 of climate change has wreaked havoc on Caribbean  |SIDS| . In 2004, Hurricane Ivan, a category 3 system, wi
mpact of climate change and provide resources for  |SIDS|  to respond to those challenges. The recently conc
ean energy. Just a few days ago, on 27 September,  |SIDS|  demonstrated their leadership, by creating the fi
the first-ever intergovernmental organization for  |SIDS| , the Small Island Developing States Sustainable E
 Developing States Sustainable Energy Initiative ( |SIDS|  DOCK). The statute came into force with a members
inable energy and climate-resilient organization.  |SIDS|  DOCK is intended to provided a collective voice, 
this first-ever intergovernmental organization of  |SIDS| . 15-29946 11/33  A/70/PV.28 03/10/2015 On behalf
.28 03/10/2015 On behalf of the States members of  |SIDS|  DOCK, we wish to express our gratitude to our par


contexts = get_context_from_keyword(
    documents=df[df['year'] == 2015]['text'],
    window_width=50,
    keyword='pv', random_seed=42
)
for x in contexts:
    print(x)

reed Sustainable 10/33 15-29946  03/10/2015 A/70/ |PV| .28 Development Goals (SDGs). Beyond the immediate
ental organization of SIDS. 15-29946 11/33  A/70/ |PV| .28 03/10/2015 On behalf of the States members of 
lic of China has 12/33 15-29946  03/10/2015 A/70/ |PV| .28 brought about major improvements in our countr
g and United Nations reform. 15-29822 5/29  A/70/ |PV| .26 02/10/2015 Two years ago in this very forum (s
0/2015 Two years ago in this very forum (see A/68/ |PV| .22), the Philippines outlined the five pillars th
ly engaged in the 6/29 15-29822  02/10/2015 A/70/ |PV| .26 United Nations system. Indeed, the partnership
ersations on the much-needed 15-29822 7/29  A/70/ |PV| .26 02/10/2015 and long-overdue reforms to make th
iation to the United Nations 15-29822 3/29  A/70/ |PV| .26 02/10/2015 Secretary-General for his leadershi
onal institutions 4/29 15-29822  02/10/2015 A/70/ |PV| .26 such as ASEAN should be be further strengthene
ne refugee issue 26/54 15-29562  30/09/2015 A/70/ |PV| .19 in accordance with resolution 194 (III) and th


tf_idf_df = tf_idf(
    df=df,
    tokens_column='bi_grams',
    segment_columns = None,
    min_frequency_corpus=20,
    min_frequency_document=20,
)


ax = tf_idf_df.sort_values(by='frequency', ascending=False).head(30)[['frequency']].plot(kind='barh', width=0.99)
ax.set_title("Frequency of Bi-Grams")
ax.set_xlabel("Frequency")
ax.invert_yaxis();


ax = tf_idf_df.head(30)[['tf-idf']].plot(kind='barh', width=0.99)
ax.set_title("TF-IDF of Bi-Grams")
ax.set_xlabel("TF-IDF")
ax.invert_yaxis();


tf_idf_per_year = tf_idf(
    df=df,
    tokens_column='bi_grams',
    segment_columns = 'year',
    min_frequency_corpus=3,
    min_frequency_document=3,
)


stop_words = ['twenty-fifth anniversary', 'twenty-five years', 'twenty years', 'twenty fifth']
tokens_to_show = tf_idf_per_year.query('year == 1970').reset_index()
tokens_to_show = tokens_to_show[~tokens_to_show.token.isin(stop_words)]
tokens_to_show = tokens_to_show[['token', 'tf-idf']].set_index('token')
tokens_to_show = tokens_to_show.to_dict()['tf-idf']
plot_wordcloud(tokens_to_show);


stop_words = ['seventieth anniversary']
tokens_to_show = tf_idf_per_year.query('year == 2015').reset_index()
tokens_to_show = tokens_to_show[~tokens_to_show.token.isin(stop_words)]
tokens_to_show = tokens_to_show[['token', 'tf-idf']].set_index('token')
tokens_to_show = tokens_to_show.to_dict()['tf-idf']
plot_wordcloud(tokens_to_show);


stop_words = ['seventieth anniversary']
#tokens_to_show = tf_idf_per_year.query('year == 2015').reset_index()
tokens_to_show = tf_idf_per_year.reset_index()
tokens_to_show = tokens_to_show[~tokens_to_show.token.isin(stop_words)]

# filter for "climate"
tokens_to_show = tokens_to_show[tokens_to_show['token'].str.contains('climate')]
tokens_to_show


tokens_to_show = tokens_to_show[['token', 'tf-idf']].set_index('token')
tokens_to_show = tokens_to_show.to_dict()['tf-idf']
plot_wordcloud(tokens_to_show);


keyword_count_over_time = count_keywords_by(
    df=df,
    by='year',
    tokens='tokens',
    keywords = ['nuclear', 'terrorism', 'climate', 'freedom'],
)
keyword_count_over_time.head()


ax = keyword_count_over_time.plot(kind='line')
ax.set_title("Keyword count over time")
ax.set_ylabel("# of Occurrences")

Text(0, 0.5, '# of Occurrences')


keyword_count_over_time = count_keywords_by(
    df=df,
    by='year',
    tokens='bi_grams',
    keywords = ['climate change', 'human rights', 'middle east'],
)
keyword_count_over_time.head()


ax = keyword_count_over_time.plot(kind='line')
ax.set_title("Keyword count over time")
ax.set_ylabel("# of Occurrences")

Text(0, 0.5, '# of Occurrences')


contexts = get_context_from_keyword(
    documents=df[df['year'] <= 1980]['text'],
    window_width=50,
    keyword='human rights', random_seed=42
)
for x in contexts:
    print(x)

e United Nations and the Universal Declaration of  |Human Rights|  and a threat to international peace and security.
y of the adoption of the Universal Declaration of  |Human Rights| . Bahrain is looking forward to participating ii t
Bahrain attaches great importance to safeguarding  |human rights|  and fundamental freedoms for all peoples. It is s
laim to be among those who laid the foundation of  |human rights| , while at the same time they support racist regim
y them their fundamental freedoms and inalienable  |human rights| . 114. Colonialism and apartheid are crimes agains
ity and would be compromised. 183. In the area of  |human rights|  there is already talk of a so-called third genera
s already talk of a so-called third generation of  |human rights| . At the same time we must note with dismay that i
eas of the world not even the first generation of  |human rights| —that is, those basic rights flowing from the conc
lementation of the so-called second generation of  |human rights| , that is, those basic rights which flow from the 
s of the world. All those refugees are victims of  |human rights|  violations, both those who seek asylum as the res


keywords = [
    'terrorism', 'terrorist', 'nuclear', 'war', 'oil',
    'syria', 'syrian', 'refugees', 'migration', 'peacekeeping', 
    'humanitarian', 'climate', 'change', 'sustainable', 'sdgs'
]

freq_df = count_keywords_by(df, by='year', tokens='tokens', keywords=keywords)
freq_df.head()


df.groupby('year')['num_tokens'].sum().head()

year
1970    137933
1971    228954
1972    244654
1973    242385
1974    258090
Name: num_tokens, dtype: int64


# compute relative frequencies based on total number of tokens per year
freq_df = freq_df.div(df.groupby('year')['num_tokens'].sum(), axis=0)
# apply square root as sublinear filter for better contrast
freq_df = freq_df.apply(np.sqrt)
freq_df.head()


plt.figure(figsize=(10, 3))
sns.set(font_scale=1)
sns.heatmap(
    data=freq_df.T, 
    xticklabels=True, yticklabels=True, cbar=False, cmap="Reds"
)
sns.set(font_scale=1);

	# of Non-Nulls	% Nulls	% Zeros	Mean	St Dev.	Coef of Var	Skewness	Kurtosis	Min	10%	25%	50%	75%	90%	Max
session	7,507	0.0%	0.0%	49.6	12.9	0.3	-0.2	-1.1	25	31.0	39.0	51.0	61.0	67.0	70
year	7,507	0.0%	0.0%	1,994.6	12.9	0.0	-0.2	-1.1	1,970	1,976.0	1,984.0	1,996.0	2,006.0	2,012.0	2,015
text_length	7,507	0.0%	0.0%	17,967.3	7,860.0	0.4	1.1	1.8	2,362	9,553.8	12,077.0	16,424.0	22,479.5	28,658.2	72,041
num_tokens	7,507	0.0%	0.0%	1,480.3	635.2	0.4	1.1	1.7	187	793.6	1,005.5	1,358.0	1,848.0	2,336.4	5,688
num_bi_grams	7,507	0.0%	0.0%	588.5	243.6	0.4	1.0	1.6	58	321.0	408.0	544.0	726.0	912.0	2,185

	# of Non-Nulls	% Nulls	Most Freq. Value	# of Unique	% Unique
country	7,507	0.0%	ALB	199	2.7%
country_name	7,507	0.0%	Albania	199	2.7%
speaker	7,507	0.0%	<unknown>	5,429	72.3%
position	7,507	0.0%	<unknown>	114	1.5%
text	7,507	0.0%	33: May I first convey to our [...]	7,507	100.0%
tokens	7,507	0.0%	['may', 'first', 'convey', 'pr[...]	7,507	100.0%
bi_grams	7,507	0.0%	['first convey', 'albanian del[...]	7,507	100.0%

	session	year	country	country_name	speaker	position	text	text_length	tokens	num_tokens	bi_grams	num_bi_grams
0	25	1970	ALB	Albania	Mr. NAS	<unknown>	33: May I first convey to our President the congratulations of the Albanian delegation on his election to the Presidency of the twenty-fifth session of the General Assembly?\n34.\tIn taking up the...	51419	[may, first, convey, president, congratulations, albanian, delegation, election, presidency, twenty-fifth, session, general, assembly, taking, work, agenda, twenty-, fifth, session, general, assem...	4092	[first convey, albanian delegation, twenty-fifth session, general assembly, twenty- fifth, fifth session, general assembly, twenty-fifth anniversary, united nations, peace-loving member, member st...	1510
1	25	1970	ARG	Argentina	Mr. DE PABLO PARDO	<unknown>	177.\t : It is a fortunate coincidence that precisely at a time when the United Nations is celebrating its first twenty-five years of existence, an eminent jurist so closely linked to our Organiza...	29286	[fortunate, coincidence, precisely, time, united, nations, celebrating, first, twenty-five, years, existence, eminent, jurist, closely, linked, organization, elected, preside, general, assembly, b...	2341	[fortunate coincidence, united nations, first twenty-five, twenty-five years, eminent jurist, closely linked, general assembly, argentine government, excellency mr, mr president, felicitous choice...	884

	frequency
token
nations	124508
united	120763
international	117223
world	89421
countries	85734
peace	72625
development	62632
states	59713
people	59338
security	58425
economic	53833
must	51880
also	47403
new	44453
country	40878
assembly	40413
community	38294
government	37430
general	37138
organization	36434

	year	token	frequency	tf-idf
1286	1970	political climate	5	16.98
4895	1971	political climate	5	16.98
6127	1971	international climate	3	9.91
6579	1972	political climate	15	50.94
7822	1972	international climate	7	23.12
...	...	...	...	...
207523	2015	climate finance	3	20.16
207525	2015	fight climate	3	20.16
207821	2015	climate system	3	18.57
209036	2015	new climate	3	13.05
209374	2015	international climate	3	9.91

Citation¶

Setup¶

Exploratory Data Analysis¶

Dataset Summary¶

Impurity¶

Examples¶

Explore Non-Text Columns¶

Explore Text Column|¶

Top Words Used¶

Distribution of Text Length¶

Word Frequency¶

TF-IDF¶

Per Year - 1970¶

Per Year - 2015¶

Keywords in Context¶

Bi-Grams¶

By Year¶

Count Keywords Over Time¶

	terrorism	terrorist	nuclear	war	oil	syria	syrian	refugees	migration	peacekeeping	humanitarian	climate	change	sustainable	sdgs
year
1970	0.01	0.01	0.04	0.05	0.00	0.01	0.01	0.01	0.00	0.02	0.01	0.01	0.02	0.00	0.00
1971	0.01	0.01	0.03	0.05	0.01	0.01	0.01	0.03	0.00	0.01	0.02	0.01	0.02	0.00	0.00
1972	0.04	0.02	0.04	0.05	0.01	0.01	0.01	0.01	0.00	0.00	0.01	0.02	0.02	0.00	0.00
1973	0.02	0.01	0.04	0.05	0.01	0.01	0.01	0.01	0.00	0.00	0.01	0.02	0.03	0.00	0.00
1974	0.01	0.01	0.04	0.04	0.03	0.01	0.01	0.01	0.00	0.01	0.01	0.02	0.03	0.00	0.00

	nuclear	terrorism	climate	freedom
year
1970	192	7	18	128
1971	275	9	35	205
1972	393	379	74	280
1973	347	124	79	266
1974	478	24	71	316

	climate change	human rights	middle east
year
1970	0	143	190
1971	0	119	323
1972	0	124	282
1973	0	200	365
1974	0	127	458